I'm conducting a MANOVA on my data. I have two independent variables (gender and diagnosis grouping). However, my DVs failed on tests of univariate normality and the assumption of equality of covariance matrices was violated. Is there anything I can do or any alternatives I could use instead?
Hi Hannah
Yes - don't use MANOVA. In fact the only time I have ever seen it used is in psychology. Just do a straightforward linear regression for each dependent variable. Don't forget the assumption of normality is on the residual, not the dependent variable.
Or you could look at PERMANOVA (https://en.wikipedia.org/wiki/Permutational_analysis_of_variance).
Dodgy versions available in R, fully tested software from Primer-E (www.primer-e.com).
Dear Hannah
In MANOVA analysis, if the dependent variables are very high correlated with each other then there may be simpler way of combining the variables.Then, carrying out ANOVA.
In addition, it is not recommended to apply MANOVA if you have many independent variable to combine. Factor analysis is very useful in this problem.
I hope this will help,
Zuhair
- Badges
- Science topic